AvatarMachine Ai

Announcing FractalGPT - an ecosystem of LLM and reasoning models, self-evolving, in competition with #OpenAi #ChatGPT and #GPT4 vc.ru/future/635360-…

🔥 Firefunction-v2, new open-weights function-calling model🔥 I'm super excited to announce Firefunction-v2, our latest open-weights! - Competitive with GPT-4o at function-calling - 1/10 of GPT-4o cost and 2x the speed - Retains both conversation and function-calling capabilities - Available on both Huggingface and Fireworks API

@arankomatsuzaki Aran, you need 100% only. But research is cool. We achieved 100% on MultiArith dataset math problems with FractalMath

Transformers Can Do Arithmetic with the Right Embeddings Achieves up to 99% accuracy on 100 digit addition problems by training on only 20 digit numbers with a single GPU for one day repo: github.com/mcleish7/arith… abs: arxiv.org/abs/2405.17399

@omarsar0 Bad model, benchmarks leaks to training set

Phi-3 Technical Report Microsoft presents a new 3.8B parameter language model called phi-3-mini. It's trained on 3.3 trillion tokens and is reported to rival Mixtral 8x7B and GPT-3.5. Has a default context length of 4K but also includes a version that is extended to 128K (phi-mini-128K). Combines heavily filtered web data and synthetic data to train the 3.8B models. It also reports results on 7B and 14B models trained on 4.8T tokens (phi-3-small and phi-3-medium). phi-3-mini achieves 69% on MMLU while phi-3-small and phi-3-medium achieve 75% and 78% on MMLU. The authors claim that because of the limited size of this model, it has less capacity to store "factual knowledge" making it weaker for certain tasks. This is something that can be resolved using an external search engine.

@omarsar0 Also please check our report for YNDX.family community, it is Yandex employees, about FractalMath - multi-agent system for solving math reasoning problems. Outperforms ChatGPT in this task with SOTA 100%. youtube.com/watch?v=TXyoTn…

Multi agents - that’s where everything is headed. DM me if you are working on it.

@omarsar0 Hi, we have created FractalGPT multi-agent system: we have own MASL, and math theory for agents self-organised behavior, self-evolving. 15 apr: webinar about our research, with demos (RAG and FractalMath) and metrics. youtube.com/watch?v=Zz59xt…

Вышел наш доклад в сообществе YNDX Family! За 30 минут очень подробно рассказали о возможностях FractalMath - ИИ агентов в решении задачек, которые даже ChatGPT тянет нестабильно. Материал подготовлен на основе онлайн-встречи YNDX Family youtube.com/watch?v=TXyoTn…

Introducing FractalGPT multi-agent Ai technology capable to build cooperative, decentralised Ai landscape. Check video with step by step historical Ai agents review, our agentic RAG system, metrics of efficiency youtube.com/watch?v=Zz59xt…

@slashML We do research in the file of Multi-agent Ai, it is cross-discipline, works with NLP, CV and other MLs

ML researchers who are not in NLP, what are you researching? Please share. reddit.com/r/MachineLearn…

@VitalikButerin @beffjezos We at FractalGPT make multi-agent Ai, MAS, with self-evolving agents, logic reasoning. It is some kind of decentralized Ai

Does biomedical acceleration count in your view? $100m to Balvi x.com/vitalikbuterin… $100m to Kanro x.com/vitalikbuterin… $336m book value, not sure how much they cashed out, to Methuselah Foundation prnewswire.com/news-releases/… d/acc themed, both cyber/info defense tech and biomedical, is my primary philanthropic interest at the moment. Maximal open source on all three, including the biomedical (check the balvi receipts). I plan to continue!

@Harada_TEKKEN Where's Eddie Gordo? You promised he'd be out in early March and he's still not here!

TEKKEN8 Release date: January 26th, 2024 鉄拳8 2024年1月26日 発売! 4K Youtube version Link: youtube.com/watch?v=fzqvmF… #TEKKEN8 #鉄拳8 #TEKKEN #鉄拳

@Harada_TEKKEN Where's Eddie Gordo? You promised he'd be out in early March and he's still not here!

I'm deeply shocked to hear that. Inomata-Sensei used to come to my office and play games with me. She was gone too soon. Rest in peace sensei… ———- Illustrator and Animator Mutsumi Inomata Passed Away animecorner.me/?p=126650

@Lomashuk Cool. We at FractalGPT on the way to truly autonomous agents, not based on prompting. I just got back from Tbilisi yesterday, where I gave a report on our multi-agent system that solves math problems accurately at the OpenTalks.ai AI conference.

Happy to share our thoughts on Cybernetic Economy combining AI with blockchain that will help make the world more open and safe.🚀

We at FractalGPT is moving toward AGI developing multi-agent framework and SDK for autonomous agents. More: fractaltech.ru

@arankomatsuzaki LoL. similar to our research at fractalGPT

Google presents Self-Discover The proposed method improves LLM performance on reasiong tasks It outperforms Self-Consistency by more than 20%, while requiring 10-40x fewer inference compute arxiv.org/abs/2402.03620

@omarsar0 Curious why FractalGPT is not here. FractalMath multi-agent Ai system has been accepted on OpenTalks international conference, with SOTA accuracy 100% on MultiArith dataset, outperforming ChatGPT with ANY prompting technique (arithmetic reasoning). opentalks.ai/en

LLM based Multi-Agents Discusses the essential aspects of LLM-based multi-agent systems. It includes a summary of recent applications for problem-solving and word simulation. It also discusses datasets, benchmarks, challenges, and future opportunities to encourage further research and development from researchers and practitioners.

FractalMath multi-agent Ai system has been accepted on OpenTalks international conference, as first with SOTA accuracy 100% on MultiArith dataset, outperforming ChatGPT with ANY prompting technique (arithmetic reasoning). opentalks.ai/en/timetable20…

@arankomatsuzaki Aran, do you know that FractalMath multi-agent Ai system has been accepted on OpenTalks international conference, as first with SOTA accuracy 100% on MultiArith dataset, outperforming ChatGPT with ANY prompting technique (arithmetic reasoning). opentalks.ai/en/timetable20…

DeepSeekMath The 7B open LLM performs nearly on par with GPT-4 on MATH benchmark arxiv.org/abs/2402.03300

@omarsar0 With multi-agent system FractalMath achieve 100% accuracy on Multiarith dataset, beating ChatGPT with 97%

Paper: arxiv.org/abs/2402.00157

LLMs for Mathematical Reasoning Introduces an overview of research developments in LLMs for mathematical reasoning. Discusses advancements, capabilities, limitations, and applications to inspire ongoing research on LLMs for Mathematics.

@omarsar0 Elvis, do you know that FractalMath has been accepted on OpenTalks Ai conference, as first multi-agent system for Arithmetic reasoning. I assume you should add it to review. Here agenda: opentalks.ai/ru/timetable#!…